Sentence object stops at non-spaced colons?

There seems to be something wrong with inner (<a-i>s) or outer (<a-a>) sentence selection in Kakoune. The sentence object stops at colons even when not followed by a space. For example, positioning the cursor on the letter, w in one.two.three and typing <a-i>s will select two., as if it were a sentence ā€“ which obviously it is not.

Could anyone on the list confirm? I am using the Kakoune binary available on Debian testing/unstable.

This happens to me as well. Sentences are harder than you think, since there may be truncated words that feature space after dot, like ā€œGood day, Ms. Smith.ā€

Emacs by default expects that each sentence ends with two spaces after the dot

Yes, of course, there is the two-space (US) versus one-space convention after the dot that ends a sentence. If we follow some of the conventions in the US, dot + space may signal an abbreviation rather than sentence end.

But the problem here is that Kakoune stops at dot, even at dot + nonspace, which cannot be a sentence ending and cannot be an abbreviation either (even on the US convention). This just seems wrong to me. The stopping regex should not be:

[.;:!?]

but rather:

[.;:!?]\s

(where the space may be horizontal or vertical).

(By the way, the choice of end-of-sentence characters might be a Kakoune option worth considering; some users may want to exclude the semi-colon, for example ā€“ but this is another issue).

Addendum:

Actually, dot + space can also signal the end of an abbreviation (rather than sentence ending) in French, Spanish, Portuguese, etc. So stopping at dot + space will sometimes catch abbreviations rather than sentence endings, but this is better than stopping any dot, which catches a) sentence endings, b) abbreviations, c) the initial dot of ... [what we should catch is the last of the 3 dots, the one before the space], and d) any word with dots inside, such as email adresses and many standard functions in the R statistical language, etc.

2 Likes